Information and Technology for Better Decision Making
2006 Survey of Reserve
Component Spouses
Administration, Datasets, and Codebook
Additional copies of this report may be obtained from:
Defense Technical Infonnation Center
ATTN: DTIC-BRR
8725 John J. Kingman Rd., Suite #0944
Ft. Belvoir, VA 22060-6218
Or from:
http://www.dtic.mil/dtic/order.html
Ask for report by
DMDC Report No. 2006-030
March 2007
2006 SURVEY OF RESERVE COMPONENT
SPOUSES:
ADMINISTRATION, DATASETS, AND CODEBOOK
Defense Manpower Data Center
Survey & Program Evaluation Division
1600 Wilson Boulevard, Suite 400, Arlington, VA 22209-2593
Table of Contents
Acknowledgments
Defense Manpower Data Center (DMDC) is indebted to numerous people for their
assistance with the 2006 Survey of Reserve Component Spouses (2006 RCSS '), which was
conducted on behalf of the Office of the Under Secretary of Defense for Personnel and
Readiness (OUSD[P&R]). The survey program is conducted under the leadership of Timothy
Elig, Chief of the Survey and Program Evaluation Division.
Policy officials contributing to the development of this survey included: John Winkler,
Wayne Spruell, Tom Bush, Virginia Hyland, Dan Kohner, Richard Krimmer, James Scott, Col
Kathleen Woody (USAFR), and Col Nilda Urrutia (ANG), all from the Office of the Assistant
Secretary of Defense for Reserve Affairs. Other contributing officals included Jane Burke, Cathy
Flynn, and Lin Porter (Military Community and Family Policy). Howard Weiss of Purdue
University was also an important contributor.
DMDC’s Program Evaluation Branch, under the guidance of Brian Lappin, Branch Chief,
is responsible for the development of questionnaires used in the survey program. The lead
developer on this survey was Rachel Lipari. She was supported in these efforts by Lindsay
Rock, DMDC, and Megan Shaw and Kristin Olson, Consortium Research Fellows.
DMDC's Survey Technology Branch, under the guidance of James Caplan, Branch Chief,
is responsible for monitoring survey administration and survey database construction. The lead
analyst on this survey was Margaret Emma Holland, SRA International, Inc. She was supported
by Matthew Perry Consortium Research Fellow. Data Recognition Corporation (DRC)
performed data collection and editing.
DMDC’s Personnel Survey Branch, led by Richard Riemer, former Branch Chief, and
Jean Fowler, current Branch Chief, is responsible for the sampling and weighting methods used
in the survey program. Michael Paraloglou, SRA International, Inc., used the DMDC Sampling
Tool to plan the sample and developed the weights for this survey. He also used DMDC’s
Statistical Analysis Macros to calculate the estimates presented in this tabulation volume. Susan
Reinhold and Carole Massey, DMDC, and Deborah West, Northrup Grumman Corporation,
provided programming support for the sampling and weighting tasks.
IV
Table of Contents
Page
Introduction . 7
Overview of Report . 7
Method . 8
Survey Instrument . 8
Sample . 10
Respondents . 13
Survey Development and Administration . 16
Address Update Procedures . 21
Processing of Updates . 23
Survey Materials and Their Distribution . 24
Processing Returned Surveys . 27
Survey Analysis Files . 29
Estimation . 29
Data Structure . 30
Variables in the Survey Analysis Files . 3 1
Using H . 34
References . 38
Appendix
A. 2006 Survey of Reserve Component Spouses: Paper Fonn . A-l
B. 2006 Survey of Reserve Component Spouses: Web Form . B-l
C. Communications . C-l
D. Annotated Web Survey Form . D-l
E. Coding Scheme . E-l
F. Alphabetical Variable List for the Survey Analysis Files . F- 1
G. Positional Variable List for the Survey Analysis Files . G- 1
H. Frequency and Percentage Distributions for Variables in the Survey Analysis Files . H-l
I. Flat File Layout for the Basic -Release Data File . I- 1
J. Notes on Analysis . J-l
K. Examples of Analysis . K-l
L. Crosswalk of Web Instrument to Paper Instrument . L- 1
M. Crosswalk of RCSS to Previous Reserve Component Member and Spouse Surveys . M-l
v
Table of Contents
List of Tables
1 . Member Stratification Variables . 1 1
2. Factors Defining Key Reporting Domains (Member) . 12
3. Sample Allocation for the 2006 Survey of Reserve Component Spouses by Member
Characteristics . 13
4. Final Sample Relative to Drawn Sample . 14
5 . Location Rates, Response Rates, and Completion Rates . 15
6. Mailing Timeline and Return Results . 1 8
7. E-mail Address Availability by Service . 26
8. Analysis File Names . 30
List of Figures
1 . Survey Control System . 19
2. Address Updating Procedures . 22
3. The Structure of the Confidential File . 31
4. Annotated Example of a Table from G . 35
2006 SURVEY OF RESERVE COMPONENT SPOUSES:
ADMINISTRATION, DATASETS, AND CODEBOOK
Introduction
The Human Resources Strategic Assessment Program (HRSAP), Defense Manpower
Data Center (DMDC), conducts both Web-based and paper-and-pencil surveys to support the
personnel information needs of the Under Secretary of Defense for Personnel and Readiness
[USD(P&R)]. These surveys assess the attitudes and opinions of the entire Department of
Defense (DoD) community on a wide range of personnel issues. A Web-based survey program
with postal- and e-mail notification, known as the Status of Forces Surveys (SOFS), provides
data several times per year on active-duty and Reserve component members and DoD civilian
employees. Paper-and-pencil surveys with postal- and e-mail notification are used to obtain data
on sensitive topics (e.g., sexual harassment) and from populations who may have limited Internet
access (e.g., spouses of active and Reserve members).
The 2006 Survey of Reserve Component Spouses (2006 RCSS) utilized both modes of
administration — the Web as well as paper- and -pen — and was designed to assess the attitudes and
opinions of active-duty spouses on a wide-range of quality of life issues. Data were collected by
mail and Web, between November 2005 and June 2006. 1 The sample consisted of 38,549
Reserve Component spouses. A total of 1 1,001 eligible spouses returned usable surveys, which
represent an adjusted weighted response rate of 3 1 .4%.
Overview of Report
DMDC (2006a) provides details on sampling and weighting.
This report also documents the procedures used to develop the instrument, design the
sample, conduct the survey, process the data and prepare analysis weights. Along with the
survey instrument and communications to the sample members (A, B and C, respectively), the
methods section includes details on how the survey was conducted.
Following the summary of the survey methodology is a description of the survey analysis
file layout and key variables. Appendix D-M address key concepts required for the analysis of
complex survey data and the structure of records in the survey analysis files are introduced in
this section. The appendix in this report include:
• A and B - Web and paper survey instruments.
• C - Samples of all possible communications sent to sample members during the
survey administration: letters, emails, and brochure.
1 The initial survey field period closed February 9, 2006. There were 3,091 spouses incorrectly flagged as
population ineligible during the original field period. DMDC elected to re-open the field from May 1- June 1, 2006
to give them an opportunity to participate.
7
• Conventions for variable naming and construction are provided in D (annotated
questionnaire) and E (coding scheme),
• F, G, and H list the names and values of all variables in the basic-survey dataset and
the Privacy-Act confidential variables.
— F lists the variables in alphabetic order and flags the Privacy-Act confidential
variables with an asterisk (*).
— G lists the variables in the order that they appear in the dataset. Variables with
the same function are grouped together, (i.e., all variables used for weighting are
located together).
— H provides a frequency for each variable with the SAS values, OS flat file
values and SAS labels in the order that the variables appear in the dataset. In
addition to the variables available on the basic-survey file, H contains details for
the confidential variables that had to be suppressed to preserve the privacy of
survey respondents and nonrespondents.
• I provides the record layout for the basic-survey flat file.
• The SAS code used to construct the analytic variables are included in J.
• Examples of analyses are provided in K.
• F and M lists all questionnaire items and identifies where they have been used in
previous DMDC surveys of active-duty members or spouses.
Method
Survey Instrument
A copy of the 2006 RCSS Web and paper questionnaires is provided in A and B. The
survey was subdivided into the following seventeen topic areas:
1 . Background Information — Member’s active-duty background and total years of
military service; spouse military ID card and enrollment in DEERS; and spouse
characteristics, including education, personal goals, race/ethnicity, U.S. citizenship,
English as a second language, age, and personal experiences with the military.
2. Housing — Distance to nearest military installation and problems in gaining access to
installation.
2 SAS® is a registered trademark of SAS Institute Inc., Cary, NC, USA.
3 The OS flat fde is a text version of the dataset. The variables are in the columns and the records are in the rows.
This data can be loaded into any statistical software package.
8
3. Your Spouse’s Activations/Deployments — Member’s time away from home and
characteristics of activations over the past 24 months, including duration, volunteer
status, deployment, and location.
4. Your Spouse’s Activations/Deployments Since September 11, 2001 — Member’s
activation and deployment history since September 11, 2001, and current
activation/deployment status; number of activations; spouse preparedness for
activation; advance notification; household income before, during, and following
activation; health and dental care coverage; impact of activation on financial well¬
being; number of hours spouse worked; number, length, and location of deployment s);
means of communicating with member; coping; member’s post-deployment behavior;
support services received by member; spouse preparation for member’s return;
difficulty of spouse’s readjustment to member’s return; interaction with military point
of contact; and military-provided support for families.
5. Effect of Deployments on Children — Emotional/behavioral impact of deployment on
children and ways children cope with deployments.
6. Activation/Deployment Expectations — Reduced and extended lengths of
activation/deployment, member leaving sooner than expected, and member having less
time between activations/deployments than expected.
7. Preparedness — Spouse’s awareness of and access to important family documents
during deployments, as well as financial steps taken to prepare for deployments.
8. Feelings About the National Guard/Reserves — Overall satisfaction with the National
Guard/Reserve way of life; support for member’s participation; factors affecting
spouse support; member’s National Guard/Reserve career plans; likelihood of
activation/deployment in the next year; and spouse commitment to member’s staying
in the National Guard/Reserve.
9. Marital History — Military membership at time of marriage, years married, satisfaction
with marital relationship, and change in frequency of problems with relationship over
past year.
10. Children and Legal Dependents — Number of children and legal dependents living
inside or outside the home.
1 1 . Child Care — Use of child care, primary source of child care, monthly cost of child
care, days of work missed because of lack of child care, and the impact of child care
issues on member staying in the National Guard/Reserve.
12. Elder Care — Number of elderly family members receiving care from spouse, whether
any elderly family members live with spouse, and amount of care giving required.
13. Employment — Spouse employment status and history, reasons for not looking for
work, hours worked per week, reasons for working part-time, characteristics of
principal employment, and shiftwork.
9
14. Financial Well-Being — Financial goals, contributions of member’s National
Guard/Reserve and spouse’s income to total household income, financial problems
experienced, saving habits, and general financial condition.
15. Health and Well-Being — Perceptions of stress and social support.
16. Programs and Services — Use of, and satisfaction with, military-provided programs and
services; the most likely way to leam about and use support programs and services; use
of Military OneSource and the primary reason for not using it; use of, and satisfaction
with, TRICARE programs and reasons for not using them; and comparisons between
TRICARE and civilian health and dental plans.
17. Communicating with You (About Survey) — Preference for Web versus paper surveys
and reasons for not completing the survey on the Web.
Sample
The target population for the 2006 RCSS consists of spouses of Reserve component
members from the Selected Reserve in Reserve Unit, Active Guard/Reserve (AGR/FTS/AR;4
Title 10 and Title 32), Individual Mobilization Augmentee (IMA) programs from the Army
National Guard (ARNG), U.S. Army Reserve (USAR), U.S. Navy Reserve (USNR), U.S.
Marine Corps Reserve (USMCR), Air National Guard (ANG), and U.S. Air Force Reserve
(USAFR), who (1) have at least six months of service at the time the questionnaire is first fielded
and (2) are below flag rank. In addition, at the time of the survey, for the spouse to remain
eligible they must have indicated being currently married to a Reserve component member. A
Reserve component member married to another Reserve component member would be eligible
for the survey depending on their spouse’s status, not their own. The sample consisted of 38,549
individuals; 1 1,001 ultimately provided usable survey responses.
Constructing the Frame and Drawing the Sample
The sample frame was constructed from DMDC’s March 2005 Reserve Components
Common Personnel Data System (RCCPDS) and July 2005 Defense Enrollment Eligibility
Reporting System (DEERS) Medical Point-in-Time Extract (PITE) if the spouses were also
eligible for benefits. The actual source infonnation for constructing the sampling frame and
identifying key domains consisted of a computer accessible file totaling 418,276 spouse records.
The sample drawn from the sampling frame consisted of 38,549 individuals. Table 3 presents a
summary of the sample allocation by Service.
Stratification Variables
The frame was stratified (divided into mutually exclusive population groups) for
sampling using the six variables listed in Table 1.
4 Names for this program vary among Reserve components: AGR/FTS/AR is a combination of Active
Guard/Reserve (AGR), Full-Time Support (FTS), and Active Reserve (AR).
10
Table 1.
Member Stratification Variables
Dimension of Stratification
Levels
Active Status During Prior 24 Months
Not active in prior 24 months
Active SOC in prior 24 months
De-activated in prior 23 months
Reserve Component
Army National Guard
US Army Reserve
US Naval Reserve
US Marine Corps Reserve
Air National Guard
US Air Force Reserve
Reserve Program
TPU/Unknown
AGR 10
AGR32
Military Technicians
IMA
Paygrade Group 7
E1-E3
E4/Enlisted Unknowns
E5-E6
E7-E9
W1-W5
01-03/0fficer Unknown
04-06
Race/Ethnic Category
Non-minority
Minority
Gender
Male/Unknown
Female
Researchers identified population subgroups of particular interest to policy officials.
These reporting domains were defined using the demographic variables shown in Table 2.
Multiple versions of most of these variables were created to permit varying levels of detail for
analysis and reporting.
The sample size and allocation were detennined using the DMDC Sample Planning Tool
(Deever & Mason, 2002). The Tool uses a formal mathematical procedure (Chromy, 1987) to
determine the minimum cost (i.e., minimum size) allocation that meets precision requirements
(e.g., ± 5 percentage points) imposed on prevalence estimates for key reporting domains.
11
Table 2.
Factors Defining Key Reporting Domains (Member)
Factor
Levels
Active Prior 24 months and De-active Special Not active in prior 24 Months
Operations Code Prior 23 months
Active SOC in prior 24 Months
De-activated in prior 23 Months
Active Special Operations Code on Prior 13
No ASOC 13-24 Months
to 24 Months
ASOC 13-24 Months
Active Special Operations Code on Prior 12
No ASOC 1-12 Months
Months
ASOC 1-12 Months
Active Special Operations Codes on Prior 24 Noble Eagle
Months
Enduring Freedom
Iraqi Freedom (SOFR0309)
Reserve Component
U.S. Army National Guard
U.S. Army Reserve
U.S. Naval Reserve
U.S. Marine Corps Reserve
Air National Guard
U.S. Air Force Reserve
Component
Reserves
National Guard
Pay Grade Group
E1-E3
E4
E5-E6
E7-E9
W1-W5
01-03
04-06
Program
TPU
AGR 10
AGR32
MILTECH
IMA
Race-ethnic
Non-minority
Non-Hispanic Black
Hispanic
Other Race
Within each stratum, the sample was selected with equal probability and without
replacement. Sampling rates varied across the strata, so individuals were not selected with equal
12
probability overall. Table 3 presents a summary of the sample allocation for the total population
and by gender, paygrade group, race/ethnicity, geographic region, and family status by Service.
Table 3.
Sample Allocation for the 2006 Survey of Reserve Component Spouses by Member
Characteristics
Sample
Total
Army
National
Guard
Army
Reserve
Naval
Reserve
Marine
Corps
Reserve
Air
National
Guard
Air Force
Reserve
Total
38,549
6,700
7,105
5,570
7,141
6,749
5,284
Activated/Deactivated
Not active in prior 24 months
22,108
3,350
3,936
4,465
2,407
4,148
3,802
Active SOC in prior 24 months
515
100
56
25
191
111
32
De-activated in prior 23 months
15,926
3,250
3,113
1,080
4,543
2,490
1,450
Activated/Not Activated
No Active SOC lto 24 months
22,106
3,349
3,936
4,465
2,406
4,148
3,802
Active SOC 1 to 24 months
16,443
3,351
3,169
1,105
4,735
2,601
1,482
Race Ethnic Category’ 2
Unknown
1,087
73
98
360
304
106
146
Non-minority
24,423
4,148
3,757
3,247
4,710
5,071
3,490
Minority
13,039
2,479
3,250
1,963
2,127
1,572
1,648
Gender
Male
31,209
5,887
5,443
4,386
6,574
5,255
3,664
Female
7,340
813
1,662
1,184
567
1,494
1,620
Pay Grade Group
E1-E3
3,377
458
710
501
1,135
273
300
E4/Unknown
8,962
1,888
1,873
1,118
899
1,689
1,495
E5-E6
7,978
1,295
1,650
1,092
1,373
1,853
715
E7-E9
3,750
384
842
229
1,026
785
484
W1-W5
1,328
720
249
62
297
0
0
01-03
5,607
1,055
989
1,156
332
956
1,119
04-06
7,547
900
792
1,412
2,079
1,193
1,171
Reserve Program
Unknown
1,101
267
144
409
66
116
99
TPU
28,923
5,370
5,978
4,178
5,213
4,654
3,530
AGR/TAR
3,609
562
473
948
819
682
125
Military Technicians
2,442
501
261
0
0
1,297
383
IMA
2,474
0
249
35
1,043
0
1,147
Note. Counts for unknowns are may not be included.
13
Respondents
Sample Losses
The original sample file contained 38,549 records. Losses to the drawn sample are listed
in Table 4 and reviewed here. Sample members were lost from the sample for three main
reasons: (1) self-reported or other ineligibility for the survey, (2) an inability to locate the sample
member, and (3) refusal to participate in the survey or other failure to respond to the survey.
A total of 2,340 sample members (9.31%) were lost from the final sample through
classification as ineligible. Elimination of ineligibles resulted in decreasing the sample to
90.69% (N=34,961) of its original size.
Table 4.
Final Sample Relative to Drawn Sample
Sample counts
Weighted estimates of
population
n
%
n
%
Drawn sample
38,549
418,267
Ineligible on master files
-1935
5.02%
-19,170
4.58%
Self-reported ineligible
-1653
4.29%
-17,560
4.20%
Total: Ineligible
-3,588
9.31%
-36,730
8.78%
Eligible sample
34,961
90.69%
381,537
91.22%
Not located (estimated ineligible)
-164
0.43%
-1408
0.34%
Not located (estimated eligible)
-1,226
3.18%
-11,819
2.83%
Total not located
-1,390
3.61%
-13,227
3.16%
Located sample
33,571
87.09%
368,310
88.06%
Requested removal from survey mailings
-138
0.36%
-1,561
0.37%
Returned blank
-220
0.57%
-2,157
0.52%
Skipped key questions
-981
2.54%
-12,508
2.99%
Did not return a survey (estimated ineligible)
-2508
6.51%
-23,505
5.62%
Did not return a survey (estimated eligible)
-18,723
48.57%
-197,363
47.19%
Total: Nonresponse
-22,570
58.55%
-237,094
56.68%
Usable responses
11,001
28.54%
131,216
31.37%
In general, spouses’ residential addresses were used as the primary addresses of choice,
followed by the members’ residential addresses. In cases where residential addresses could not
be identified, however, member unit addresses were used. Procedures used to locate spouse of
Reserve Component members are explained in a later section that describes the Survey Control
System. Because of this address update procedure, less than 3.61% of the drawn sample (1,390
of 38,549) was lost because the sample members could not be located. Personnel records for this
14
group had missing, incomplete, or out-of-date addresses, and steps designed to obtain complete,
current addresses for these records were unsuccessful.
Losses attributable to either ineligibility or unlocatability resulted in a sample that was
87.09% of the drawn sample. Individuals in this remaining sample may be further categorized as
nonrespondents versus respondents. Nonrespondents included the following groups: sample
members who contacted the operations contractor (by mail, fax, e-mail, Web, or telephone) and
asked to have their names removed from the survey mailing list, and 21,231 sample members
who did not return a survey.
Respondents included all sample members who completed on the Web 50% of applicable
questions5. Respondent also needed to answer the two critical questions that determined
eligibility (your marital status and your spouse military status). At the conclusion of the survey
fielding, 11,001 eligible, locatable sample members had returned usable surveys
Location, Response and Completion Rates
Beginning in 1995, DMDC standardized its methods for calculating response rates and
completion rates using procedures patterned after those advocated by the Council of American
Survey Research Organizations (CASRO). CASRO noted that varying operational definitions of
response rates can lead to problems or confusion (e.g., when awarding contracts requiring pre¬
specified response rates or when interpreting the results of a survey). As a result, CASRO
formed a task force to recommend guidelines for standardizing the operational definitions of
response rates. The new DMDC procedures closely follow CASRO ’s Sample Type II design
(see Council of American Survey Research Organizations, 1982).
Table 5 provides location, response, and completion rate information. The location rate
is defined as the proportion of eligible sample members that were located. The completion rate
is defined as the proportion of the located sample that returned usable surveys. The response rate
is defined as the proportion of eligible sample members that returned usable surveys.
Table 5.
Location Rates, Response Rates, and Completion Rates
Observed Operational Rates
Weighted Operational Rates
Location rate for eligible
96.2%
96.7%
Completion rate for eligible
35.4%
38.1%
Response rate for eligible
34.1%
36.8%
5 Applicable questions are those to be completed by all respondents and excluded items that could be skipped over
depending on prior answers.
15
Survey Development and Administration
The 2006 RCSS continues a line of research on active-duty spouses begun with the 1985
DoD Surveys of Officer and Enlisted Personnel and Military Spouses. In 1992 and 1999,
DMDC conducted subsequent Joint Service surveys of active-duty spouses. Many key topics
covered by the 2006 RCSS were also included in its predecessors; however, questions have been
updated, expanded, or streamlined in the 2006 RCSS. The survey was administered by both
Web and paper- and -pencil questionnaires. Although both surveys largely covered the same
content, the Question numbering differed. Both survey fonns are in Appendix A.
The survey was hosted on the operations contractor’s secure Web site so that sample
members could complete the survey online. At the entry point to the survey, sample members
were prompted for their personal ticket number to gain entry to the survey. The Privacy Notice
and a page of frequently asked questions (FAQ’s) were linked from here.
The survey allowed respondents to return to the previous page or move to the next page.
In addition, buttons located below the last Question on each page allowed the respondent to clear
their response(s) or save and exit the survey. Questions were answered by clicking on radio
buttons, check boxes or by making a choice from a drop-down list. The respondent could change
answers or could save, exit, and return at another time to change answers. The final page had
another “Save and Exit” button and a “Done” button, both with full text explanation of their
functions.
For those people who had not completed the questionnaire on the Web system, we mailed
the paper fonn to sample members along with the third reminder send on December 19th 2005
(see Table 6 for more information on the mailings).
Survey Administration
The survey administration process began in November 2005, with the mailout of
notification letters to sample members (minus original ineligibles). The original field period was
November 7, 2005, through February 9, 2006. Up to three additional postal communications
were mailed to sample members throughout this field period. The survey field was re-opened on
May 1, 2006, in order to communicate with 3,566 sample members originally misclassified as
ineligible. During the May field period, a postal notification and one postal reminder were sent.
The field closed on June 1, 2007.
In addition, sample members a valid with e-mail address on record could have received
an e-mail notification plus up to eight e-mail reminders during the November-February field
period. During the May field period, sample members with a valid e-mail address could have
received an e-mail notification and up to four e-mail reminders. Postal and e-mail mailings
stopped once the sample member returned a survey.
May fielding ticket miss match
Identified Issue: DRC associated incorrect Web Ticket Numbers for the re-field when
preparing the postal Notification letter, dated May 1, 2006. This was discovered by DRC
16
approximately two weeks after the Notification letters were mailed. The mismatch resulted in
Web Survey returns with incorrect Ticket Numbers.
Subsequent postal communications for the re-field provided correct Ticket Numbers. All
e-mail communications for the re-field provided correct Ticket Numbers.
Process: After discussions with DMDC, the following screener question was presented
prior to the first survey question. This was posted May 18, 2006 at 1 1 :06 AM CDT.
Did you access this survey using the Ticket Number from a postal letter dated
May 1,2006?
Yes (value =1)
No (value = 2)
If the Next Page button was selected without answering the question, respondents
received a reminder pop-up that it had not been answered. After one reminder, respondents were
allowed to advance to Question 1 of the survey.
Analysis: There were 179 Web survey returns (165 complete and 14 partial.)
• Using the screener question, the survey submit date (SRDATE), and presence of an e-
mail address, DRC determined the correct data match for 159 Web returns.
• There were 20 returns that the source could not be identified. These returns were
designated as .B in the dataset.
• The variable created for these returns was MIS MTCH where
- . = not mis-matched
- 0 = Corrected Ticket
- 1 = Recodes Un-Matchable
17
Table 6.
Mailing Timeline and Return Results
Print File
Number
Creation
Mail Drop
Number
of
Mailing Numbers and Groups
Date*
Date
Sent
PNDs
Notification Domestic
10/28/05
11/7/05
32,631
2,051
Notification Foreign
10/28/05
11/7/05
72
21
Notification Domestic Reminder 1
11/22/05
11/21/05
748
122
Notification Foreign Reminder 1
1 1/22/05
11/21/05
26
8
Subtotal: Notification
33,477
2,202
Reminder 1 Domestic
11/28/05
12/1/05
31,128
1,459
Reminder 1 Foreign
11/28/05
12/1/05
96
49
Reminder 1 Domestic Re -mail 1
12/6/05
12/7/05
706
108
Reminder 1 Foreign Re -mail 1
12/6/05
12/7/05
76
43
Subtotal: Reminder 1
32,006
1657
Reminder 2 Domestic
12/9/05
12/14/06
28,389
793
Reminder 2 Foreign
12/9/05
12/14/06
162
96
Reminder 2 Domestic Re -mail 1
12/16/05
12/19/06
305
51
Subtotal: Reminder 2
12/16/05
12/19/06
25
13
Reminder 3 Domestic
28,881
953
Reminder 3 Foreign
12/19/05
1/5/06
27,112
597
Reminder 3 Domestic Re -mail 1
12/19/05
1/5/06
148
72
Reminder 3 Foreign Re -mail 1
1/10/06
1/11/06
662
102
Reminder 3 Domestic Re -mail 2
1/10/06
1/11/06
34
13
Subtotal: Reminder 3
1/16/06
1/17/06
10
4
Reminder 4 Domestic
27,966
788
Reminder 4 Foreign
1/19/06
1/26/06
23,181
170
Reminder 4 Domestic Re -mail 1
1/19/06
1/26/06
84
10
Reminder 4 Foreign Re -mail 1
1/30/06
1/31/06
168
2
Subtotal: Reminder 4
1/30/06
1/31/06
4
0
May 2006 Notification Domestic
23,437
180
May 2006 Notification Foreign
4/25/06
5/1/06
3,518
272
Subtotal: May 2006 Notification
4/21/06
5/1/06
33
23
May 2006 Reminder 1 Domestic
3,551
295
May 2006 Reminder 1 Foreign
5/10/06
5/15/06
3,481
250
May 2006 Reminder 1 Domestic Re-mail 1
5/10/06
5/15/06
33
12
May 2006 Reminder 1 Domestic Re-mail 2
5/22/06
5/23/06
134
5
Subtotal: May 2006 Reminder 1
5/22/06
5/23/06
7
1
3,658
268
*Print file creation date: This is the date records were identified for inclusion in the mailing and written to a print
file.
18
Survey Control System
The Survey Control System (SCS)6 was used to monitor the data collection process and
to track all data transactions over the course of the survey administration. The datasets in the
SCS include sample members’ names and addresses, but do not contain data obtained from the
survey instruments. Because of privacy concerns, SCS datasets are not available for basic
release.
The operations contractor uses the SCS to store and update project data, monitor
mailings, respond to documents returned as postal non-deliverables (PNDs), and detennine
survey participation and eligibility status. The SCS consists of five datasets: the ORIGDAT file,
the ADDRESS file, the MASTER file, the HISTORY file, and the MAILING file. Figure 1
displays the relationships among those datasets.
Figure 1.
Survey Control System
ORIGDAT file. The ORIGDAT file consists of 38,549 records, one record for each
member of the sample. It is the original sampling frame file sent to the operations contractor by
DMDC. The original file is loaded onto the operations contractor’s computer system and
converted to a SAS dataset. As the file was converted into a SAS dataset, the SCS generated a
6The SCS refers to the set of data files as well as the program or operating system which maintains those files.
19
unique identification number (INRECNO) for each record. This number identifies the sample
member throughout the SCS and also in returns data sets, comment text files and other specify
text files. The names and some demographic data from the ORIGDAT file were loaded into the
MASTER file in preparation for the first mailing. The addresses from the ORIGDAT file were
loaded into the ADDRESS file.
ADDRESS file. The ADDRESS file tracked the postal and e-mail addresses that were
maintained for each sample member. The ADDRESS file contains one record for each postal
and address for each sample member (e.g., if there were five addresses located for one sample
member during the survey administration, that sample member has five separate records in the
ADDRESS file) yielding an ADDRESS file containing 164,819 records. Each record is uniquely
identified by the combination of INRECNO (identifying the sample member) and an address
number (ADDRNO) assigned to each address. This address number is the sequential order of
receipt of the address for a particular sample member. For example, if a sample member has one
address record in the ADDRESS file, the address number for that record is one. If the sample
member faxed in a change of postal or e-mail address or a credit bureau forwarded an updated
postal address for that sample member, the new address was added as address number two. The
ADDRESS file was initially loaded with postal and e-mail addresses from the ORIGDAT file.
Each record in the ADDRESS file includes the sample member’s INRECNO, address, the source
of the address, and address priority code, a variable indicating whether the record is the highest
priority address for this sample member, and variables indicating whether the address
successfully reached the sample member.
The priority code assigned to a given address number for a sample member was used to
determine the “best” or “highest priority” address for the sample member at any given time. It
was originally determined by the source of the address. Address updates obtained directly from
a sample member received a priority number of one. The order of priority of address sources
from “highest priority” to “lowest priority” is as follows, respectively:
1. updates directly from a sample member (call, fax, e-mail, Web update or letter)
2. address corrections from the U.S. postal service (ACS [electronic address change
service], ACRs [address correction requests], and ODFs [out-of-date-forwarded mail])
3. NCOA-updated addresses
4. credit bureau-updated addresses
5. DEERS residential addresses
6. DEERS unit addresses
MASTER file. The MASTER file is used by the SCS to select records for upcoming
survey mailings. This file includes a record for each member of the sample and was initially
created by extracting data from each record in the ORIGDAT file. Each MASTER record
includes the sample member INRECNO and the address number for the highest priority postal
and e-mail address in the ADDRESS file for this sample member. The MASTER file
accommodated data updates through an automated process (e.g., updating the address number in
20
use after the receipt of a postal or e-mail nondeliverable or Web update) or manual key entry
(e.g., updating infonnation in response to a telephone call, fax, letter return or e-mail from a
sample member). As new information was received for a particular record (including changes to
the highest priority address), the SCS updated the MASTER record (N=38,549) and wrote the
old record to the HISTORY file. The MASTER file also contains a set of variables which
summarize the sample member’s mailings status.
HISTORY file. The HISTORY file is a chronicle of the changes that occurred to the
MASTER file. Each HISTORY record is a subset of an outdated MASTER record with the
addition of a date and time stamp as the record is updated. That is, a HISTORY record is created
when there is a name, address, paygrade, or eligibility status change in the MASTER file. Thus,
the HISTORY file contains as many observations as there are updates to the MASTER file.
MAILING file. The MAILING file tracked all survey mailings (postal and e-mail). This
file contains one record for either an item postal mailed or e-mailed during the survey
administration or for tracking postal address updates from credit bureaus (N= 15 3, 2 19). Each
MAILING record includes the INRECNO, address number used, date of mailing, mailing status,
type of mailing, and the mailing identification code (MIC).
Address Update Procedures
Initial Address Updates
Prior to the first mailing, the operations contractor ensured all domestic residential
addresses were formatted to conform to U.S. Postal Service standards. Once the addresses were
standardized, they were sent to an outside vendor where they were checked against the National
Change of Address (NCOA) database. The NCOA software updated the address records (in
standardized format) based on change-of-address cards filed with the U.S. Postal Service. The
updated NCOA address file was returned to the operations contractor and integrated into the
SCS. The NCOA-updated addresses were added to the ADDRESS file and became the current
ADDRNO with the “highest priority code assigned” in the MASTER file.
After the NCOA-updated data was added to the SCS, another file was compiled of
sample members who had an incomplete address or an address identified by NCOA as an
undocumented move (i.e., the sample member had moved, but NCOA did not have a new
address). The operations contractor sent copies of this file to three credit bureaus (Experian,
Trans Union and CSC Credit Services)7 to detennine whether a complete, up-to-date address for
these sample members could be found. The results were integrated into the SCS, updating
records in the ADDRESS file.
Ongoing Address Updates
Address update procedures also occurred when (a) additional address records were
received after NCOA processing, (b) a survey document was returned as undeliverable, (c) a
7Experian, Trans Union and CSC Credit Services are outside vendors with consumer-credit information databases.
Social security numbers of sample members with incomplete or out-of-date address information were forwarded to
the vendors for address updates when the mailing dataset contained no valid address.
21
sample member self-reported a name, rank, or address change, or (d) the U.S. Postal Service
forwarded address correction information. Figure 2 outlines these procedures.
Figure 2,
Address Updating Procedures
As a new address was entered into the ADDRESS fde, its source (NCOA, credit bureau,
postal Address Correction Requested card, telephone call, fax, letter, Web, or e-mail) was
recorded and a new address number was assigned. The priority assigned to the address was
based upon the source of the update and the date and time of the address (see the description of
priority, for the ADDRESS file). At any given time, the current address used corresponded to
the address number with the highest priority code.
22
If all known addresses for a sample member were returned Postal Non-Deliverable Mail
(PND), the sample member’s record in the MASTER file was flagged “no address available.”
All “no address available” records were forwarded to the three credit bureaus. The credit
bureaus returned files containing addresses for each submitted record, with the date on which the
credit bureau received the address. If more than one address for a sample member was received
from credit bureaus, the address number corresponding to the address with the most recent
receipt date received the highest priority code. If one or more of the credit bureaus returned a
previously unattempted address, the MASTER and ADDRESS files were updated and a re -mail
was sent to the sample member. If none of the vendors had an updated address for the sample
member, the operations contractor designated the sample member “nonlocatable” and stopped
further mailings.
Processing of Updates
Updates from Sample Members
Sample members could provide an updated address in a variety of ways. Updates from
sample members could be communicated via the toll-free telephone number (either by speaking
to the operations contractor’s Call Center staff or by leaving a voice mail message). In addition,
sample members could mail, fax, e-mail or the survey Web site all updated infonnation was
entered into the SCS. Updates made on the Web site were loaded directly into the SCS before
the start of the survey; once the survey fielding period stared, the Web update page was no
longer available. Other updates were entered into the SCS by the operations contractor’s Call
Center staff by the close of business on the day following receipt of the update.
Updates from the U.S. Postal Service
There are several types of address updates provided by the postal service. They are
detailed below; each includes a description of the processing steps.
1 . Postal Non-Deliverable Mail (PND): The sample member moved and no forwarding
address was available. The mail piece was returned to the operations contractor. The
operations contractor removed the letter from the envelope and scanned it to capture
the Mailing Identification Code (MIC) in the lower right corner. A file of the MICs
was loaded to the SCS so the records could be updated as PND. This was done as
necessary to coincide with the mailing/re-mailing schedule. If sample member had
another address on file (e.g., the unit address), that address was used for the next
mailing for the next mailing. If no alternate address was on file, the Social Security
Number was sent to the credit bureaus in search of a new address.
2. Address Change Service (ACS; electronic): About six weeks prior to the first mailing,
the operations contractor applied to the postal service for the ACS. The postal service
assigned a participant code, which was placed in the address block of the letter. The
operations contractor requested semi-weekly files, which the postal service provided
on diskette via Express Mail. The operations contractor loaded the files upon receipt
or before another mailing was prepared.
23
3. Address Correction Requests (ACR; hard-copy): The outbound envelopes contained
the endorsement “Address Service Requested.” The post office provided the
corrections via hard copy cards that were sent to the operations contractor. The
corrections were entered into the SCS by the operations contractor’s Call Center staff,
typically by close of business the next day but no later than prior to the preparation of
the next mailing.
Survey Materials and Their Distribution
Each eligible sample member received at most four original mailings: a notification
letter and brochure explaining the survey program, a reminder letter, a reminder letter with a
paper survey and a third reminder letter. The notification and reminder letter mailings contained
a letter, except for the second reminder which contained a letter, paper survey and business reply
envelope. All letters included information about using the Web as an option to complete the
survey.
In addition, e-mail was used to communicate with sample members. Not every sample
member had an e-mail address. However, those sample members for whom we had an e-mail
address received an e-mail announcement and up to eight e-mail reminders. Samples of the
letters and e-mail communications are provided in C.
General Mailing Procedures
Prior to every mailing, the SCS searched the records in the MASTER file to identify
which records should be excluded (e.g., sample members self-reported as ineligible for survey
participation, sample members who had already returned survey forms, and members with no
valid addresses available). For re-mails (sent between mailings), the SCS identified only those
records that had been updated since the prior mailing. More specifically, the SCS identified
records that had resulted in PNDs or had been manually flagged for re-mailing (e.g., in response
to a sample member calling the operations contractor stating she or he had received a
reminder/thank you letter but had not received a survey, etc.).
Once all records for a particular mailing or re-mailing were identified, the SCS processed
the records based on whether the mailing would include a brochure and/or a survey fonn. If the
mailing group was large enough to lead to a cost savings from sorting, the records were run
through Group 1 postal software to sort the records according to first-class presort postal
regulations. After this procedure, a unique Mail Identification Code (MIC) was assigned to each
record. The MIC was assigned either from the survey litho code list if a survey form was sent or
independently if only a letter was sent.
Ticket Numbers for Web Survey Access
Prior to the first mailing, a list of ticket numbers for Web survey access was randomly
generated. One secure ticket number was assigned to each sample member and remained linked
to that member for the duration of the project. That is, while a member’s MIC or lithocode
changed with each mailing as described previously, the member’s ticket number did not change.
s Ticket numbers are eight alpha numeric characters generated at random.
24
The member’s unique ticket number was printed (along with the survey URL) in each letter and
e-mail sent to that individual. A member could not access the Web survey without using his or
her ticket number.
Description of Letters
Letters were printed with the record’s unique MIC listed in the address field and on the
lower right corner of the letter. If the mailing included only letters (no brochures or survey
forms), the letters were folded and machine inserted into window envelopes and sent by first
class mail. Mailings that included a brochure or a survey followed the same procedure through
the letter printing process. The MIC on the cover letter was used to pair the letter with the
correct enclosure. During the matching process, ten percent of the mailing was visually checked,
comparing numbers printed on the letter with the brochure or survey number for quality control.
Any mismatched pairs initiated further investigation of the matching process. This procedure
ensured that each brochure or survey was sent to the person designated to receive it. Depending
on the sample size, the letters and matched enclosures were machine or hand inserted into
envelopes, metered if necessary, and sent by first class mail.
The status of each mailing was tracked throughout the data collection so that address-
correction information could be incorporated into all relevant mailings. When a mail piece came
back PND, the next mail piece was sent to a new address (if one could be obtained during the
mailing period). For all mail pieces that came back PND, re-mails were completed if a
newer/updated address could be found.
DMDC provided the operations contractor with the text, letterhead and signature for the
cover letters. The letters explained why the survey was being conducted, how the survey
information would be used, and why participation was important. See C for copies of the letters.
The letters were approved and printed on letterhead from the office of the Under Secretary of
Defense and signed by the Under Secretary of Defense (Personnel and Readiness), David S.C.
Chu. The letterhead and signature were printed in blue, and the text and recipient infonnation of
all letters were printed in black. In addition to including a name and address (which was also
used as the mailing information for the window envelopes), each letter included a personalized
salutation. The salutation addressed each sample member by his/her gender. For example, a
letter to an Active Duty spouse would have included the salutation, “Dear Mrs. Smith.”
Mailouts
Table 6 lists the mailing dates and return results for each of the mailouts and re-mailings.
For the main notification mailing, sample members were sent a letter and brochure notified
sample members that they were selected for this survey and encouraged their participation. The
notification letter was mailed to 32,703 sample members on November 7, 2005.
The first reminder letter was sent to 3 1,224 sample members on December 1, 2005. The
letter, thanked sample members for completing the survey if they had done so, and reminded
them to complete the survey if they had not. The second reminder letter was sent to 28,55 1
sample members on December 14, 2005. The letter again thanked sample members for
25
completing the survey if they had done so, and reminded them to complete the survey if they had
not.
The third reminder mailing provided sample members the option to complete a paper
survey. For this mailing, a letter, paper survey and a folded business reply envelope were
provided. The survey packet was mailed to 27,260 sample members on January 5, 2006.
The fourth postal reminder letter was sent to 23,265 sample members on January 26,
2006. The letter thanked sample members for completing the survey if they had done so, and
reminded them to complete the survey if they had not.
The field was re-opened on May 1, 2006, as stated earlier. The second notification
packet was sent to sample members initially flagged as ineligible and offered sample members
the option to complete the survey on paper or on a secure Web site. For this mailing, a letter,
paper survey, brochure and business reply envelope were provided. This packet was mailed to
3,551 sample members on May 1, 2006.
A reminder letter was sent to 3,5 14 sample members on May 15, 2006. The letter
thanked sample members for completing the survey if they had done so, and reminded them to
complete the survey if they had not.
E-mail was used to communicate with sample members. E-mail addresses were
purchased from an outside vendor. The outside vendor maintains a customer database of e-mail
addresses that has been lawfully collected and compiled from consumers pursuant to a notice that
advised them that their personal data was being collected. Table 7 below shows the percent of
sample members by Service for whom at least one valid e-mail.
Table 7.
E-mail Address Availability by Service
Army
National
Guard
Army
Reserve
Navy
Reserve
Marine
Corps
Reserve
Air
National
Guard
Air Force
Reserve
Total
Valid address available
14%
15%
14%
14%
15%
14%
14%
No valid address available
86%
85%
86%
86%
85%
86%
86%
Sample members with e-mail addresses received at most an e-mail notification and eight
reminders. Table 8 lists the e-mail dates and e-mail addresses bounced. E-mail addresses
“bounced” identifies the address was invalid at the time DMDC attempted contact. This is
analogous to a postal PND. E-mail address “sent” is not the same as e-mail received. It is
analogous to the non-PND return experienced during a mailed survey. It is not known if the mail
was delivered to the intended individual, only that it was not returned.
26
Table 8.
E-mail Communication Timeline
Communication Type
E-mail Drop Date
Number Sent
Number Bounced
Notification
11/21/05
6486
10215
Reminder 1
11/28/05
4467
41
Reminder 2
12/6/05
3833
20
Reminder 3
12/14/05
3336
23
Reminder 4
12/22/05
3054
15
Reminder 5
12/28/05
2928
24
Reminder 6
1/11/06
2693
27
Reminder 7
1/19/06
2439
27
Reminder 8
1/27/06
2306
0
May Notification
5/1/06
129
41
May Reminder 1
5/5/06
123
0
May Reminder 2
5/11/06
90
0
May Reminder 3
5/17/06
86
0
May Reminder 4
5/23/06
81
0
Processing Returned Surveys
Once a respondent completes the survey, data are stored in an indexed file on the Web
(data) server. Web and paper survey returns are merged into one dataset. Paper survey returns
require additional work to input the data (explained below). Prior to providing each dataset to
DMDC, the operations contractor copied the indexed file to their internal network using FTP
protocol. The data are then converted to a sequential fonnat, and the validate program reads and
loads the data to the dataset.
All paper returned surveys were logged in and opened by the operations contractor upon
receipt. If the envelope contained the survey booklet and other materials (e.g., extra comments,
photographs, non-relevant items), the operations contractor separated it from the survey.
Bundles of this type of correspondence (white mail) were sent to DMDC by regular surface mail
or FedEx ground after all surveys were received. If the white mail appeared to be urgent, the
operations contractor contacted DMDC to detennine how it should be handled.
Survey booklets were batched for image scanning and assigned a batch number. The
booklets were separated by pages, stacked in page/booklet, and forwarded for scanning. As the
surveys were scanned, the batch number and a serial number (unique to each survey) were
printed on each page of the survey.
The surveys were machine-edited for light marks, multiple marks, and alignment.
Damaged forms were repaired, if possible, and scanned with non-damaged forms. If it was not
possible to scan the documents, they were batched separately and key-entered.
27
Regardless of the mode of survey submission, the operations contractor processed all
survey information according to DMDC approved administration plans and coding schemes.
DMDC Coding Scheme
To convert the raw data into the item scores that appear in the data files (basic-release
and confidential files), DMDC provided the operations contractor with an annotated copy of the
survey form (see D) and the coding notes (see E). Every attempt is made to capture all
information from completed surveys and preserve the data so that secondary analysts can later
create variables that were not anticipated by DMDC researchers. To accomplish these goals,
DMDC subscribes to a variety of coding conventions (see D).DMDC uses “forward” coding
when coding inconsistent answers in items with skip patterns. Data on the starting Question
accepted as marked and data for the items within the skip pattern are edited to be consistent with
the starting question. However, an unedited version of each item is preserved in a confidential
dataset.
For Web respondent, the coding scheme is used to “smart skip” respondent. This does
not allow respondents to view questions that they have indicated with previous answers do not
apply to them. For example, if a respondent indicated on question 18 (SR018= 1) that they had
not ever tried to go to a military installation becoming a spouse of a National Guard/Reserve
since then they did not see Question 19, which asks “Since becoming a National Guard/Reserve
spouse, have any of the following caused you problems in gaining access to your nearest military
installation?” Only those with the affirmative answer; yes, I have tried to go to a military
installation are shown the questions.
Coding or Keying Open-Ended Items
The Web survey contained twenty-one open-ended items. The original text responses
from these items were captured verbatim into a SAS® data set that is linked by the unique
identification to the survey data. The paper fonn had fewer open-ended items. The operations
contractor keyed all verbatim. Text data in the SAS® files for open-ended items were spell¬
checked. Identifiers (e.g., proper names, addresses, e-mail addresses, phone numbers, locations,
or social security numbers) were replaced with generic terms.
Fifty-Record Check
After receiving the 5% of returned records, the operations contractor ran a “50-record
check.” This is a check to verify that the coding scheme and skip patters are working. DMDC
checked the resulting data to detennine if there were any unanticipated problems in the coding
procedures (e.g., respondents were consistently answering in an unexpected manner). Minor
corrections to these procedures were necessary as a result of this check and were reviewed by
DMDC prior to production of the initial SAS® dataset. At the completion of the 50-record
check, the operations contractor compiled the full set of returned surveys. The data were then
cleaned and edited following the coding scheme.
28
Survey Analysis Files
This section (a) provides an overview of requirements for analysis of the data,
(b) documents the structure of survey analysis fdes created for the 2006 RCSS survey,
(c) describes the assembly of the analysis files, and (d) provides an overview of the variables in
the survey analysis files.
Estimation
Analysis of this data requires use of weights to compensate for the unequal selection
probabilities and to account for differential nonresponse among population subgroups. The
analytic weights were poststratified to population totals so that weighted sample estimates would
reflect population values.
In general, the procedures used to compute sample estimates of population parameters
(including population totals, means, proportions), tests of hypotheses, regression relations, and
their associated variances are derived from the probability structure that gives rise to the
observations. As with other surveys involving complex probability structures, most of the
parameter estimates of interest in this survey take the form of non-linear statistics. Examples
include domain means and proportions where the denominator values are unknown and must be
estimated from the sample data. The estimator takes the form of a ratio of random variables (i.e.,
the ratio of the estimated numerator and denominator totals or counts). In general, ratio
estimates are not unbiased and their variances cannot be expressed in closed form. The variances
are, therefore, approximated. The bias in a ratio estimate depends on the variance associated
with the denominator total or count and can usually be ignored in samples having a large number
of observations. As a working rule, the bias may be assumed negligible if the number of
observations on which the estimate is based exceeds 30 or is otherwise large enough so that the
coefficient of variation [SE(x)/x] of the denominator is less than .10 (cfi, Cochran, 1977, pp.
153-165).
Two common variance estimation methods for complex sample data are linearization
(Taylor series approximation) and replication. Wolter (1985) provides a detailed discussion on
methods used for variance estimation from sample surveys, including Taylor series
approximation and replication methods.
Many of the standard statistical software packages, such as SPSS9 and older versions of
SAS,10 compute variance estimates only for simple random samples. Using standard statistical
programs with the appropriate eligibility indicator (ELIGFLGW) and the analytic weight
(FINALWGT) to analyze this data will produce accurate point estimates, but variance estimates
will not account for the complex sample design. Variables have been included in the analysis file
so that Taylor series estimates can be computed for a stratified without replacement design, using
either SUDAAN9 or the recently available SAS Survey Procedures.
9 SPSS® is a registered trademark of SPSS Inc., Chicago, IL, USA.
10 SAS added survey procedures in Version 7, expanding them in releases 8.0 and higher.
29
Data Structure
Care was taken in the preparation of the survey analysis files to provide basic access to
data from the survey with sufficient information for accurate estimations, while meeting
requirements for participant and non-participant anonymity. As described below, some detailed
variables have been deleted from the basic-release files either because (a) they provide too great
a chance of identifying an individual or (b) they are not needed to analyze the survey data. For
the latter reason, some demographic variables are available on basic files only in a collapsed
version. In addition to a basic-release file, a confidential file (containing a more complete set of
variables than the basic-release file) has been prepared for internal DMDC use. Files were
prepared as SAS and SPSS system files. An ASCII (Operating System or OS) flat file was
prepared from the basic-release SAS system file. File names are indicated in Table 9.
Table 8.
Analysis File Names
Type of File
File Name
Basic-release File - SAS
Confidential File - SAS
Basic-release File - SPSS
Basic -release File - OS
RCSS06B.7BDAT
RCSS06C.7BDAT
RCSS06B.POR
RCSS06B.DAT
The structure of the confidential file is shown in Figure 3. The confidential file contains
the basic-release file plus additional confidential variables. All variables in the confidential file
are documented in this report. Appendix F and G list all variables with a notation to indicate
which variables are confidential and show where each variable is documented. Intermediate
weighting variables that appear only in the confidential file are documented by DMDC (2006a).
Variables that appear in collapsed fonn in the basic-release part of the file and in a fuller version
only in the confidential file are discussed later.
Analyses
Both the confidential file and basic-release file contain 38,549 records, one for every
sampled. As depicted in Figure 3, these records can be divided into 3 subgroups. The
Nonrespondents subgroup, includes all records indicated by ELIGFLGW=3, where no usable
response was received or ineligibility could not be determined (27,548).
Assignment of a record to the other two subgroups was based on whether (a) an
individual returned a “completed” survey; and (b) the individual was eligible for the survey .
Final eligibility was limited to those in both the March 2005 Reserve Components Common
Personnel Data System (RCCPDS) and the July 2005 Defense Enrollment Eligibility Reporting
System (DEERS) Medical Point-In-Time Extract (PITE) who did not contact the operations
contractor to indicate that they were ineligible.
30
The analytic dataset should consist of records for the Known Self- or Proxy- reported
Ineligibles and Eligible Respondents subgroups. Both the Eligible Respondents (ELIGFLGW=1)
and Known Self- or Proxy-reported Ineligibles (ELIGFLGW=2) are included because both types
of records were used for poststratification to population totals; both types of records are needed
to compute accurate variance estimates by Taylor series linearization. To analyze the eligible
completed responses use the analytic weight, FINALWGT, subset the file to ELIGFLGW = 1,2
(i.e., records with non-zero weights), and restrict the subpopulation for analysis to
ELIGFLGW=1.
Figure 3.
The Structure of the Confidential File
Confidential and
Detailed
Methodological
Subgroups Basic-release File Variables
Eligibility Flag
Value and Number
of Records
ELIGFLGW=3
n= 23,690
ELIGFLGW=2
n= 1,653
ELIGFLGW=1
n= 11,001
Note. The shaded portion represents the subset of the data typically required for analysis.
Variables in the Survey Analysis Files
Basic-survey Dataset
The variables in the basic-survey dataset fall into five categories: (1) Information
gathered on the survey, (2) Variables constructed for analysis, (3) Infonnation on operations,
(4) Infonnation from sampling and record data, and (5) Information on weighting. Variables are
grouped in these categories in G and H.
Information gathered on the survey. These variables came directly from the survey or
were constructed using only information from the survey. There is at least one variable for every
item in the survey except for a few items that had to be removed to preserve confidentiality. The
annotated questionnaire (see Appendix D) contains the item names, the values used to code the
pre-specified alternatives, and references to applicable coding notes in E.
DMDC uses a standard naming convention for most variables. In general, the survey-
derived variables can be classified as variables that begin with either “SA,” “SR,” or “X.” The
31
naming of “SR” variables is reviewed using the example variable, “SR052A.” For the 2006
Survey of Reserve Component Spouses, variables names begin with “SR” to denote the
population (active-duty spouse) and the survey administration year. The following three
numbers correspond to the questionnaire item number. For example, the third through fifth
digits indicate the main Question number (046), the sixth digit typically indicates the sub-
Question item, such as (in this example) item A from a list of items in Question 46.
The “SR” variables are a set of primarily demographic items that are identically named
across all DMDC surveys. The “SR” serves as a mnemonic for self-report with the remainder of
the name indicating the data being collected. For example, “SRRACE” is the variable name for
the item that asks sample members what race they consider themselves to be. Although all
survey data are self-reported, the “SR” is used to distinguish survey-reported infonnation from
DMDC-provided information (e.g., the variable “SRRACE” from the survey is differentiated
from the variable “RACE” from DMDC databases). When possible, “X” is reserved to create
special crossing (marginal) variables for key analyses. “X” variables typically involve
imputation for missing data and, like “SR” variables, are intended to be consistent across DMDC
surveys. For more information on variable naming conventions, see Appendix E.
Variables constructed for analysis . An “R” as the last letter of a variable listed in
Appendix F, G, and H is an indication that the variables may have been recoded to create special
analysis. Only one version of each variable is available in basic-dataset. For example, certain
demographic variables, including some information collected on the survey, had to be censored
to preserve the anonymity promised to survey respondents and nonrespondents. For example,
SR015R is a recoding of SR015.
Certain key demographic variables were constructed for DMDC analyses. These analytic
variables, starting with “X,” are based primarily on self-reported information from the survey.
Typically, where the self-reported information was missing on important demographics (e.g.,
Service, paygrade, location, or respondent gender) data were imputed from members’ or
spouses’ administrative record.
The race and ethnicity questions were combined to be reported in accordance with the
Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity
(1997). Also, items were combined to derive employment indicators based on U.S. Census
Bureau’s Decennial Census and Current Population Survey (2002).
Appendix J documents many of the decisions made in the analyses reported by DMDC
(2006b). For a large number of survey items, analysts must make decisions on the treatment of
special codes (such as Not Applicable.).
32
Information on operations . The DMDC-provided identification number, RCSS2006, is
unique and is used to identify responses as they are processed. Other variables are created by the
operations contractor but are too detailed to be in the basic-release file.
Information from sampling and record data. Most of the variables used in sample
design and selection are too detailed to be in the basic-release file (see the later section on
confidential variables).
Information on weighting. Derivation of weights is discussed in detail in DMDC
(2006a). See Appendix K for examples of analyses using these variables:11
ELIGFLGW
FINALWGT
VSTRAT
TOTAF
Eligibility Flag
Final Weight with Non-response and Postratification Adjustments
Variance Estimation Strata
Weighting Class Strata Totals Based on Sampling Frame Counts
Full Survey Dataset
In addition to variables on the basic-survey dataset, the full survey dataset also has five
additional categories of variables: (1) the raw version of survey items that appear in a collapsed
form in the basic-release section, (2) the raw version of key demographic variables used in
analyses that appear in a collapsed fonn in the basic-release section; (3) detailed variables
created by the operations contractor to document operations, (4) detailed variables used in
sampling, and (5) detailed variables used in weighting. Variables are grouped in these categories
in Appendix F, G and H.
Privacy Act confidential variables — survey data. This section of the full survey dataset
contains the original survey variables that had a recoded version in the basic-survey dataset. To
the extent possible, recoded versions of these variables are in the basic-release file section under
variables constructed for analysis.
Privacy Act confidential variables — analysis data. This section of the full survey
dataset contains the analytic variables constructed by DMDC. To the extent possible, recoded
versions of these variables are in the basic-survey dataset section under variables constructed for
analysis.
Privacy Act confidential variables — operations data. This section of the full survey
dataset contains operational variables created by the operations contractor. These variables are
useful for methodological studies and/or were used in determining eligibility and response status.
The identifying variables describe how the record was processed once a survey was
returned. The variables BATCH, SERIAF, and FITHO uniquely identify each returned survey.
FITHO is the lithocode scanned from the survey. BATCH and SERIAF are the codes printed on
11 Two additional variables required for SUDAAN are on the dataset: NPSTRAT, poststratification population
counts; and, PSTRATA, poststratification strata.
33
the survey during scanning to identify the scan batch number and scan order of each survey.
These numbers can be used to retrieve the paper copy of a survey for a short time after it has
been scanned (e.g., should researchers want to check electronically-stored information against
the respondent’s answer on the paper survey). DUPRET and DUPRET2 indicate the receipt of
multiple returns. DUPRET2 includes blank returns in the multiple counts; DUPRET excludes
these returns.
The classification variables describe how individual sample member’s records were
grouped and indexed. FALG FIN indicates the final disposition status of a sample member (i.e.,
survey returned, blank survey returned, not locatable, or no return). Several other classification
variables were used to categorize a survey’s final disposition. These variables are: BLKREAS,
SCSINEL, and REFUSE. BLKREAS codes the reason given by the sample member for
returning a blank survey, SCSINEL indicates the reason given by the sample member for being
ineligible, and REFUSE indicates whether a sample member refused to complete a survey.
Privacy Act confidential variables — sampling and record data. This section of the full
survey dataset contains administrative file variables and constructed variables used in
determining the sampling design. It also includes the sampling strata identifiers and counts.
Confidential variables — weighting. This section of the full survey dataset contains
variables used in analysis of non-response and in the construction of the weights.
Using Appendix H
Regardless of whether analysts use all or only portions of the database, all analysts
should replicate the results found in the tables in H. It is only by replicating these results that
analysts can be sure that they are reading the data correctly. An annotated example of an H table
is listed in Figure 4. (However, table does not reflect actual results.)
34
Figure 4.
Annotated Example of a Table from G
12006 Survey of Reserve Component Spouses
Were your parent ( s ) /guardian ( s ) in a regular Reserve Component
Service and/or National Guard/Reserve?
2SR015 3Were your parent ( s ) /guardian ( s ) in a regular Reserve
Component Service and/or National Guard/Reserve?
FREQ6
PERCENT7
OS
VALUE8
SAS
VALUE9
MEANING10
693
1 . 9
-9
.
No response
2
o
o
-8
.A
Multiple response error
23419
65.0
-1
.B
No survey return
2135
5.9
1
1
Yes, while I was growing up
3089
k D
OO
2
2
Yes, but only before I was
born
6716
18.7
3
3
No
36054
100.1
TOTALS11
12 PERCENT TOTAL DOES NOT = 100 DUE TO ROUNDING ERROR.
13G-2
SAS DATA
FORMAT NAME
TYPE
LENGTH
INFORMAT
SR079
NUM
3
STDOS2
OS DATA4
COLS
LENGTH
NA-NA
NA
1 . Codebook title and item text. The codebook title is the same for every table in
Appendix H of this codebook. It lists survey name. If applicable, the indented text
under the title presents the verbatim Question or instructions that accompany a specific
item in the survey.
2. Variable name. The variable name for a survey item is up to eight characters in
length and corresponds to the variable name that is used in the SAS(R-based, basic-
release data file. The conventions for naming survey-derived variables are
documented in Appendix E. Appendix F and G contains a full listing of the basic-
release file variables, as well as short descriptions of what the variables document.
3. Survey item text. For survey items, this text is the verbatim item wording. For other
variables, this text provides a verbal description of the variable.
4. Location of the item on the OS data file. This block provides the location of the
variable on the OS data file. The OS data block documents (a) the starting and ending
column numbers where the data are stored and (b) the number of columns that the data
occupy.
35
5. SAS data file information. This block indicates format name, variable type (character
or number), length and informat of the data in the SAS® data file. The last block
indicates the informat appropriate for reading the data from the OS data file.
6. Counts of item value responses. This column indicates the number of sample
members who fall into the category corresponding to each value for the variable. The
count provided for each variable value should correspond exactly to those that analysts
would obtain when running unweighted frequencies on all 36054 records in the
accompanying database. Before running complex statistical analyses, analysts are
encouraged to re-create these frequency tables. Re-creating the counts minimally
ensures that the data are being correctly read by the analysts’ computers and programs.
7. Respondent percentages for each value. This column indicates the percentage of
sample members who marked each variable value. The percentages are calculated by
dividing the row value in the “FREQ” column by the total listed at the bottom of the
“FREQ” column. The percentages provided for each variable value should correspond
exactly to those that analysts would obtain when running unweighted frequencies on
all 36054 records in the accompanying database.
8. Response OS values. This column presents the OS (ASCII) code for the actual or re¬
coded response values for each survey item. Further details on the values in this
column are found in either the annotated survey form or in E. For example, all
negative values are found in Appendix E.
9. Response SAS® values. This column presents the SAS® code for the response values
for each variable. Further details on the values in this column are found in either the
annotated survey form or in Appendix E. An explanation of negative values is
presented in Appendix E.
10. Explanation of the item value codes. This column presents brief verbal explanations
of the OS and SAS® coding for each survey item. If the coded infonnation
corresponds to survey response alternatives, the text in the table is the verbatim
response from the survey instrument. More detailed explanations are presented in the
annotated survey form (Appendix D) and in Appendix E.
11. Total of response frequencies and percents. The number appearing at the bottom of
the “FREQ” column is the total number of sample members in the basic-release file.
This number is the same for every table in this codebook. That is, every sample
member in the database is accounted for on every variable even if the variable
indicates only that the infonnation was missing for that sample member. The number
appearing at the bottom of the “PERCENT” column is typically 100.0. Rounding
enor, however, occasionally causes the total percentage to be slightly above or below
100.0.
12. Messages to analysts. The messages alert analysts to situations specific to a variable
including (a) rounding errors resulting in a total percentage other than 100 percent;
(b) the variable having values that are “too numerous to list;” (c) extraction of the
36
variable from another specified database; (d) creation of the variable from two or more
variables specified in the message; and (e) further clarification of the survey item
corresponding to the variable.
13. Codebook page number. This is the H page number corresponding to a specific
variable. F and G identifies the page number in H where the variable can be found.
37
References
Bureau of the Census. (2002, March). Current Population Survey: Design and methodology
(Technical Paper 63RV). Retrieved May 31, 2002, from
http://www.census.gov/prod/2002pubs/tp63rv.pdf
Chromy, J. R. (1987). Design optimization with multiple objectives. Proceedings of the
Section on Survey Research Methods, 194-199. Alexandria, VA: American Statistical
Association.
Cochran, W.G. (1977). Sampling techniques (3rd ed.). New York: John Wiley & Sons.
Council of American Survey Research Organizations. (1982). On the Definition of Response
Rates (special report of the CASRO Task Force on Completion Rates, Lester R. Frankel,
Chair). Port Jefferson, NY.
Deever, J. A., & Mason, R. E. (2002). DMDC Sampling Planning Tool (Version 2.0)
[Computer software]. Arlington, VA: DMDC.
DMDC. (2006a). 2006 Survey of Reserve Component Spouses: Statistical methodology report
(Report No. 2002-031). Arlington, V A: DMDC.
DMDC. (2006b). 2006 Survey of Reserve Component Spouses: Tabulation of responses.
(Report No. 2006-029). Arlington, VA.
OMB Bulletin No. 00-02. (2000). Guidance on aggregation and allocation of data on race for
use in civil rights monitoring and enforcement. Washington, DC: U.S. Office of
Management and Budget.
S AS Institute Inc. (2001). SAS/STAT User’s Guide, Version 8. Cary, NC.
SPSS® for Windows™ [Computer software]. (1993). Chicago, IL: SPSS Inc.
Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity, 62
Fed. Reg. 58781 (1997).
SUDD AN® Software for the Statistical Analysis of Correlated Data [Computer Software].
(1996). Research Triangle Park, NC: Research Triangle Institute.
Wolter, K.M. (1985). Introduction to variance estimation. New York: Springer-Verlag.
38
REPORT DOCUMENTATION PAGE
Form Approved
OMB No. 0704-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection
of information, including suggestions for reducing the burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports
(0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be
subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.
| PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS. j
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE
31-03-2007 Final Report
3. DATES COVERED ( From - To)
November 2005-June 2006
4. TITLE AND SUBTITLE
2006 Survey of Reserve Component Spouses - Administration, Datasets, and
Codebook
5a. CONTRACT NUMBER
5b. GRANT NUMBER
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S)
Defense Manpower Data Center (DMDC)
5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)
Defense Manpower Data Center
1600 Wilson Boulevard, Suite 400
Arlington, VA 22209-2593
8. PERFORMING ORGANIZATION
REPORT NUMBER
DMDC Report 2006-030
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES)
10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for Public Release; distribution unlimited
13. SUPPLEMENTARY NOTES
Appendices not included. Information for appendices should be directed to Timothy Elig, DMDC 1600 Wilson Blvd, Arlington, VA
22209. (703)696-5790. _
14. ABSTRACT
The 2006 Survey of Reserve Component Spouses (2006 RCSS) is part of the 2006 Survey of Military Spouses project. The 2006
RCSS was designed to measure the attitudes and opinions of spouses regarding a wide-range of issues including demographic
information, marital history, employment status, feelings about military life, use of military programs and services, and number and
age of children or other legal dependents. This report documents how the survey was fielded and the creation of reporting variables.
15. SUBJECT TERMS
Survey, Reserve Component Members, Reserve Component Spouses, Demographics, Preparedness, Children/Legal Dependents,
Health and Well Being, Employment Status, National Guard/Reserves, and Programs and Services.
| 16. SECURITY CLASSIFICATION OF:
17. LIMITATION OF
ABSTRACT
18. NUMBER
OF
PAGES
19a. NAME OF RESPONSIBLE PERSON
a. REPORT
b. ABSTRACT
c. THIS PAGE
Margaret Emma Holland
U
U
U
UU
44
19b. TELEPHONE NUMBER (Include area code)
(703) 696-4790
Standard Form 298 (Rev. 8/98)
Prescribed by ANSI Std. Z39.18